On Tue, Jun 19, 2018 at 10:39:59PM +1000, Michael Ellerman wrote:
> Ram Pai <linux...@us.ibm.com> writes:
> > Deny all permissions on all keys, with some exceptions. pkey-0 must
> > allow all permissions, or else everything comes to a screaching halt.
> > Execute-only key must allow execute permission.
>
> Another ABI change.
>
> Are we calling this a bug fix?
It is a ABI change. There are two cases where this could break an
existing application.
a) single threaded application, depending on the AMR bits of
unallocated keys to do something.
Not sure what one can achieve doing so.
b) Multithreaded application could see the difference. The scenarios is
i) Thread T2 allocates a key and associates with Memory M1
ii) Thread T1 accesses the memory M1.
Without the patch step (ii) will be successful.
With the patch step (ii) will fail.
I doubt any multithreaded applications are out there depending
on this particular behavior. And if it does, than it is
depending on a buggy behavior.
RP
